Data Analytics report

8月28日转人工率偏高诊断

UMall AI 客服截至 11:37 的转人工率诊断。

核心结论

偏高来自重复会话，不是客户增多。 截至 11:37，28日为 12/17（70.6%Source: UMall production conversation sessionsTable: production conversation session JSONL files），27日同期为 8/14（57.1%Source: UMall production conversation sessionsTable: production conversation session JSONL files）；按客户统计均为 7/12（58.3%Source: UMall production conversation sessionsTable: production conversation session JSONL files）。新增 4Source: UMall production conversation sessionsTable: production conversation session JSONL files 次全部来自重复转人工（1Source: UMall production conversation sessionsTable: production conversation session JSONL files→5Source: UMall production conversation sessionsTable: production conversation session JSONL files）。

3Source: UMall production conversation sessionsTable: production conversation session JSONL files 个客户占 8/12 次，单个取消订单占 4Source: UMall production conversation sessionsTable: production conversation session JSONL files 次。工具异常包括 afterSaleInfo 拦截 11/17 次、图片引用失败 2Source: UMall production conversation sessionsTable: production conversation session JSONL files 次。建议复用同一客户/订单的人工工单并修复附件引用；全天结束后再复核。

Source: UMall production conversation sessionsTable: production conversation session JSONL files

Scans production UMall session JSONL records, applies the same 11:37 JST cutoff to Aug 27 and Aug 28, detects transfer markers, and aggregates only de-identified counts.

同期转人工率对比
同期转人工率对比 data
DateRateDenominator
Aug 270.57Sessions
Aug 270.58Unique customers
Aug 280.71Sessions
Aug 280.58Unique customers

Sources

  1. UMall production conversation sessionsanalysis.sql

    Scans production UMall session JSONL records, applies the same 11:37 JST cutoff to Aug 27 and Aug 28, detects transfer markers, and aggregates only de-identified counts.

  2. CustomerAi after-sale status checks

    Fresh read-only checks of after-sale status for the orders represented in Aug 28 transferred sessions.

    SQL query
    python3 scripts/customer_ai_query.py afterSaleInfo --order-id <order-id>